Replace parallel STL with PPQSort.

PPQSort is supposedly quite fast: https://github.com/GabTux/PPQSort

More importantly, it does not depend on TBB fuckery, so there's no longer
a need to link with an external library that people may or may not have.

The NO_PARALLEL_STL option is out, as it was provided solely to deal with
TBB being not available. Sequential sorting is still used on emscripten.

server/TracySortedVector.hpp

@@ -102,10 +102,10 @@ public:
         const auto se = sb + sortedEnd;
         const auto sl = se - 1;
         const auto ue = v.end();
-#ifdef NO_PARALLEL_SORT
-        pdqsort_branchless( se, ue, comp );
+#ifdef __EMSCRIPTEN__
+        pdqsort_branchless( sb, se, comp );
 #else
-        std::sort( std::execution::par_unseq, se, ue, comp );
+        ppqsort::sort( ppqsort::execution::par, sb, se, comp );
 #endif
         const auto ss = std::lower_bound( sb, se, *se, comp );
         const auto uu = std::lower_bound( se, ue, *sl, comp );